Google took a significant step into the future of personal assistants on June 27, 2025, by opening early access to its new AI agent, codenamed "Project Mariner." This tool, as reported by leading tech publications TechCrunch and The Verge, is capable of autonomously performing complex, multi-step tasks directly within a web browser on behalf of the user, marking a shift from informational queries to real-world actions. Initial access to "Project Mariner" will be strictly limited and apparently granted to users in the US with a new premium subscription, tentatively called "AI Ultra." Unlike existing AI assistants that primarily answer questions or execute simple commands, "Mariner" is designed for full task completion. A user can give it a high-level command, for example: "Find and book the cheapest direct flight to Paris for next weekend and add it to my calendar." The agent would then autonomously open the necessary airline or aggregator websites, compare prices, fill out all required forms with user data, and complete the booking. This technology is based on a deep integration of the powerful Gemini language model with systems that allow the AI to "see" and semantically interact with web page elements. The launch of this project confirms the broader industry trend of moving from chatbots to full-fledged agentic systems. "Project Mariner" is a direct competitor to OpenAIs "super assistant" concept and could fundamentally change how people interact with the internet by automating routine online tasks, from shopping to travel planning and complex research.
Google Opens Early Access to "Project Mariner" AI Agent for Browser-Based Tasks
